We combine Lp quantile regression with realized measure and call it realized Lp quantile regression(Realized Lp),through incorporating measurement equation into the conventional Lp quantile regression model,in a framework analogous to Realized-GARCH.We use Realized Variance(RV)and Realized Range(RR)with different inter-vals as the employed realized measures.These are modeled in the measurement equa-tion.We find exponential power distribution allows likelihood to be developed,that facilitates pseudo-maximum likelihood estimation.We have a simulation to ascertain whether the model can estimate parameters correctly.Afterwards,we analyze empiri-cal results by using S&P 500.Both the parameters estimated and returns forecasted are analyzed.Furthermore,we also compare RV and RR with different intervals to find out which realized measures are the best.Finally,we'll compare different p in Lp quantile regression.Then more indices are used,and we find that L1.2 suits for most indices. |